This essay is a philosophical take on the Koch-Chalmers bet of 1998, which was finally settled in June 2023. It asks what the bet actually meant, and attempts to explain why Christof Koch lost, and David Chalmers won. Almost inevitably, it also focuses on what’s often called “the hard problem of consciousness”.
“It started off as a very big philosophical mystery [] But over the years, it’s gradually been transmuting into, if not a ‘scientific’ mystery, at least one that we can get a partial grip on scientifically.”
— David Chalmers [This is Chalmers’ response to a question about his bet with Christof Koch.]
“I bet consciousness will be even more baffling in 2048 than it is today. I hope to live long enough to see Koch give Chalmers another case of wine.”
— John Horgan [Source here.]
The journal Nature set the scene in this way:
“In 1998, neuroscientist Christof Koch bet philosopher David Chalmers that the mechanism by which the brain’s neurons produce consciousness would be discovered by 2023.”
Nature concluded:
“Both scientists agreed publicly on 23 June, at the annual meeting of the Association for the Scientific Study of Consciousness (ASSC) in New York City, that it is still an ongoing quest — and declared Chalmers the winner.”
[This passage can be found in an article called ‘Decades-long bet on consciousness ends — and it’s philosopher 1, neuroscientist 0'. See also note 1 on the lack of details about this bet.]
Oddly enough, this categorical victory doesn’t square (at least not perfectly) with what the science journalist John Horgan had to say on this very issue. In an article called ‘A 25-Year-Old Bet about Consciousness Has Finally Been Settled’, he writes:
“The results of the tests were inconclusive. Some data favor IIT [information integration theory]; others favor the global workspace.”
This means that from the “inconclusive” results of the “tests” on these two theories, a categorical victory still had to announced. Yet it can be said that words like “winner” and “philosopher 1, neuroscientist 0” are still too categorical (or perhaps simply theatrical) even for a bet like this.
Relatedly, some commentators have said that these two theories are “still incomplete”. (Most — perhaps all — scientific theories are incomplete. [See note 2.]) Such commentators have therefore concluded that “neuroscientists do not understand consciousness”.
A Bet About the Neural Correlates of Consciousness?
Way back in 1998, the computational neuroscientist and neurophysiologist Christof Koch had a lot of faith in finding all the neural correlates of consciousness (or of conscious states) by the year 2023. That was largely because a lot of other neuroscientists (at least those interested in consciousness) had a lot of faith in the new neuroimaging technology which had become available around the year 1998. However, even if that neuroimaging technology was indeed a godsend to neuroscientists and philosophers, all the correlations might still have had to wait a lot longer than Koch believed at the time.
It can also be assumed the philosopher David Chalmers has always happily accepted that the “the brain mechanisms underlying consciousness would [eventually be] known”. However, in 1998, he simply didn’t believe that they’d all be known by 2023.
In any case, Koch conceded that he had lost the bet.
Oddly enough, Koch then had the audacity (if that’s the right word) to make a brand new bet of a very similar kind. At least that is John Horgan’s recollection of the event. Horgan writes:
“Koch then doubled down on his bet. Twenty-five years from now, he predicted, when he will be age 91 and Chalmers will be age 82, consciousness researchers will achieve the ‘clarity’ that now eludes them. Chalmers, shaking Koch’s hand, took the bet.”
[See note 3 on the phrase “correlation doesn’t imply causation”.]
The Two Leading Theories of Consciousness
The main way of deciding this bet between David Chalmers and Christof Koch was to test the “two leading theories” of consciousness.
Those two theories are the information integration theory (IIT) and the global workspace theory (GWT).
More interestingly, advocates of the IIT tested the GWT, and the advocates of the GWT tested the IIT.
So, to be clear, IIT and GWT aren’t necessarily the two leading theories of consciousness. (It can be suspected that various idealist, “spiritualist”, sexy, etc. theories of consciousness are far more popular.) They’re the two leading scientific theories of consciousness. Indeed, even that claim needs to be qualified. IIT and GWT are the two leading theories of consciousness which are deemed — by many scientists and philosophers in the relevant fields - to be (partly or fully?) scientific. (Some idealists — such as Donald Hoffman and Bernardo Kastrup — stress the scientific nature of their own theories of consciousness.)
To repeat. It’s been said that that there are “two leading hypotheses about the neural basis of consciousness”.
Readers may now wonder how all this has been established.
It’s not being said here that they aren’t two leading (scientific) hypotheses. It’s just a question about how that was established.
Was this a numbers game?
Is it that most scientists and philosophers uphold one or the other of these two (scientific) hypotheses? (At least the ones who concern themselves with the nature of consciousness.) Alternatively, is it only that most scientists or most philosophers uphold these two hypotheses?
Some readers may also be wondering why the global workspace theory and information integration theory can’t be both true together.
So take this passage from Horgan:
“GWT explains it in terms of a set of brain structures, whereas IIT regards it as a computational function.”
These two theories (or, more simply, these two approaches) don’t seem to contradict each other… at least not as expressed in this basic form.
After all, can’t GWT’s “brain structures” be the physical embodiments (or implementations) of IIT’s “computational function[s]”? Unless, that is, the IIT approach completely disregards such (physical) embodiments or implementations. (Critics have also argued something similar about artificial-intelligence theorists and their own algorithms. [See my ‘Chalmers, Penrose and Searle on the (Implicit) Platonism and Dualism of Algorithmic AI’.])
Well, IIT doesn’t ignore the brain. That said, the emphasis on computational function[s] (as with A.I.’s emphasis on algorithms) does seem to lead in that direction.
So do GWT and IIT mutually contradict each other in many ways? Or do they do so in only some ways?
Definitions of the Word ‘Consciousness’
No one can say that “consciousness hasn’t been explained” unless they firstly tell us what consciousness is. Indeed, perhaps this issue of definitions is at least partly behind the entire “problem of consciousness”. (Stressing the importance of definitions has been classed as “pedantic”, “logic chopping” or even worse.)
These comments also impinge on the very nature of the Chalmers-Koch bet.
Specifically, one main purpose of the bet was to see (or to guess) whether or not science would have a satisfactory explanation for consciousness by 2023.
As a perfect example of the consciousness-is-an-unfinished-project position, take this passage from an article called ‘Neuroscientist loses a 25-year bet on consciousness — to a philosopher’:
“We can say, though, that being conscious involves not only having awareness of the outside world, but also having awareness of one’s self and one’s relationship with the environment.”
The author (Mo Costandi) goes on to explain this as follows:
“While the brain surely plays a major role in generating our conscious experiences, awareness of one’s self involves complex interactions between the brain and the body. In order to gain a better understanding of the full spectrum of consciousness, we will need to take these interactions into account.”
It can be strongly suspected that Chalmers himself would have a problem with bringing in the “self” and self-consciousness into a definition of “being conscious”. This won’t be because commentary on the nature of the self and on self-consciousness isn’t part of the overall story of consciousness. It’s just not a necessary part. Indeed, that passage above can be deemed to be anthropocentric. And that’s because this emphasis on the self and self-consciousness would deny many — perhaps even all — animals consciousness.
In detail.
Many philosophers stress various levels of consciousness (perhaps best highlighted in Ned Block’s 1992 paper ‘On a Confusion about a Function of Consciousness’) which seem to have a minimal role (in terms of the conscious entity discussed) for any concepts at all. (“Non-conceptual content” is also relevant here.) What’s more, some levels of consciousness seem even more basic, and therefore they’re even less likely to involve “having awareness of one’s self”. [See note 4.]
As for “having awareness of [] one’s relationship with the environment”: that seems to be even more particular when it comes to definitions of the word “consciousness” or what consciousness is. In other words, it seems to be a strong hint at the ‘extended mind thesis’, ‘embodied cognition’, ‘enactivism’ (or ‘embedded minds’), and perhaps even at the arcane philosophical notion of ‘externalism’.
Relevantly, note the words “minds”, “cognition”, “self” (i.e., not the word “consciousness”) in all these philosophical and scientific terms.
This passage from Mo Costandi [see note 5] shows us that if people define the word “consciousness” in different ways, then that will impact on their attitudes toward what consciousness is. (This is obvious, but still needs stating.)
For example, according to one definition, there is no Hard Problem at all. According to another definition, there is a Hard Problem. Indeed, according some definitions, it may be impossible to solve the Hard Problem precisely because the definitions themselves render any solution impossible. What’s more, perhaps such definitions are explicitly or implicitly (as it were) designed to make the Hard Problem hard (i.e., unsolvable).
So let’s now tackle the Hard Problem.
The Eternal Debate About the Hard Problem
Firstly, I must keep reminding myself — and readers — that the Koch-Chalmers bet wasn’t really about the Hard Problem… at least not directly. [See note 1 again on the lack of detail on this bet.]
In any case, to get a grip on the Hard Problem (see ‘Hard problem of consciousness’), it’s worth discussing the global workplace theory again and tying it directly to this problem.
Firstly, John Horgan writes:
“According to GWT, which was proposed by Bernard Baars in the 1980s, consciousness is generated by core neural architecture that processes largely unconscious information, and then the prefrontal cortex broadcasts that information to a wider network, so that it enters conscious awareness.”
The GWT theory doesn’t immediately seem to help with Chalmers’ Hard Problem. That’s primarily because it’s about how consciousness is generated. However, consciousness itself does then enter this GWT story in that the “prefrontal cortex broadcasts that information to a wider network, so that it enters conscious awareness”.
Yet Chalmers would inevitably say (if in his own words):
Sure!
But why does that information have the particular experiential nature that it does have when it’s “broadcasted”?
In terms of the ‘neural correlates of consciousness’ at the heart of the Koch-Chalmers bet.
To Koch, these correlates are essentially about
“the bits and pieces of the brain that are really essential — really necessary to ultimately generate a feeling of seeing or hearing or wanting”.
Yet Chalmers can still ask his Hard Question about these “bits and pieces of the brain”, and why they generate feelings, seeings and hearings which have the very particular phenomenal qualities (or qualia) which they do have.
In other words, no matter how many correlations are discovered between conscious states and neuronal states (indeed, even if literally all the correlations were discovered), Chalmers could still ask his Hard Question.
What’s more, Chalmers doesn’t appear to be asking for answers which refer to anything physical, functional, structural, evolutionary, etc. Or, at the very least, these things alone will never satisfy Chalmers. Thus:
(1) Chalmers isn’t asking for the physical correlates of experiences.
(2) Chalmers isn’t asking for the causal and physical connections between the brain and experiences — or any connections between experiences and anything physical.
(3) Chalmers isn’t asking for the functional (or otherwise) underpinnings of experiences.
(4) And Chalmers isn’t asking for any evolutionary answers either.
All that said, readers shouldn’t misread that list.
Chalmers is very much interested in “the easy problems of consciousness”. He’s also interested in the scientific nature and role of the neural correlates of consciousness. Indeed, after over three decades of thinking about these issues, he must also be well-versed in the neuroscience of the brain. In other words, Chalmers certainly isn’t an example of one of those philosophers of mind who shows very little interest in neuroscience and in other relevant scientific subjects. (Not that many professional philosophers fit that bill nowadays.)
That said, and as Chalmers himself is keen to stress, all the things referred to above (i.e., in the list) could be instantiated, and the experiences may still not occur. Alternatively, all these things could be instantiated, and the experiences could still be different… even very different.
Put that way, it now seems like Chalmers’ bet was very wise indeed.
Basically, Chalmers knew (or he believed he knew) that he couldn’t loose… at least when it came to his Hard Problem.
Yet on the surface at least, this bet wasn’t really about the Hard Problem. It was specifically about finding “the mechanism by which the brain’s neurons produce consciousness” by the year 2023.
So now take these words from John Horgan:
“They agreed that, for Koch to win, the evidence for a neural signature of consciousness must be ‘clear.’ That word ‘clear’ doomed Koch. ‘It’s clear that things are not clear,’ Chalmers said, and Koch, grimacing, concurred.”
These words don’t seem to be about Chalmers’ Hard Problem.
However, if all the correlations were mapped, then there would still be no (to use Koch’s term) “clarity” as to why a single neuronal state brought about the particular conscious state it does bring about. Similarly for all correlated neuronal states (or all the relevant ones) put together: clarity would still not be achieved when it comes to the Hard Problem.
Thus, Chalmers wins again.
Perhaps Chalmers wins again because his Hard Problem is insoluble on his own terms… and even on anyone’s terms.
Of course, Koch mightn’t have been referring to clarity as it’s specifically relevant to the Hard Problem. Yet if the Hard Problem is a genuine problem (i.e., if it’s not a philosophical trick), then there’ll always be a lack of clarity even for the (supposed) easy problems of consciousness.
Horgan himself also seems to argue that Chalmers believes that he knows that his own Hard Problem is insoluble.
For example, in reference to the new bet (i.e., not the original bet of 1998) between Chalmers and Koch, Horgan writes:
“‘I hope I lose,’ Chalmers said, ‘but I suspect I’ll win.’”
Horgan then concludes:
“I suspect so, too. I bet consciousness will be even more baffling in 2048 than it is today. I hope to live long enough to see Koch give Chalmers another case of wine.”
Readers may find it odd (or, as I do, highly suspicious) that Chalmers is so sure that he will “win” the bet yet again in 25 years!
Indeed, Horgan himself tuned into Chalmers’ (I will say it!) mysterianism when he says that “consciousness will be even more baffling” in 25 years.
The problem here is with the word “baffling”.
What kind of problem becomes even more baffling over time, and after more and more people have spent more and more time on it?
If the Hard Problem isn’t a genuine problem, then it’s not “baffling” at all. Perhaps there’s no solution to it because there can be no solution to it. Perhaps the Hard Problem is (metaphorically) designed to be insoluble.
It now needs to be said that the word “mysterianism” wasn’t used gratuitously a few moments ago. After all, Chalmers himself is keen to tell us the following:
“It started off as a very big philosophical mystery [] But over the years, it’s gradually been transmuting into, if not a ‘scientific’ mystery, at least one that we can get a partial grip on scientifically.”
This passage doesn’t include a concession on Chalmers’ part.
The clause “if not a ‘scientific’ mystery” means that all the (mere) correlations required may well be forthcoming by, say, 128 years and 9 months. However, the Hard Problem with remain with us… forever!
Notes
(1) It should be noted here that I haven’t been able to gain access to the relevant academic papers — for various reasons — on this bet. (There aren’t that many on it anyway.) Thus, I’ve relied on the limited details which can be found in places outside academia’s subscription policies.
In addition, the conference at which the outcome of the bet was decided included so many different philosophers and scientists, with so many different “theories of consciousness”, that it would have been almost impossible to find any relevant material. [See here.]
(2) Isn’t there a (strong) sense that many, most or even all scientific theories are incomplete? Indeed, aren’t even fundamental theories, in some weak sense at least, incomplete?
(3) One way to tie the the neural correlations of consciousness directly to the Hard Problem is the following way.
In many cases, neuroscientists do know that when brain events/states of a certain type occur, then experiences of a certain type occur.
Are these “mere correlations”?
People often — smugly? — use the phrase “correlation doesn’t imply causation”. (This phrase is almost a cliche now.) Sure, the clue is in the word “imply”. Thus, some things may well be correlated, even though there are no causal connections whatsoever between them. However, in many cases, causal connections do actually explain the correlations.
What’s more, not everyone is demanding a necessary connection between neural states/events and conscious states/events. However, there are very many well-documented and well-tested contingent connections. Yet precisely because these connections are all still contingent (i.e., not necessary), then this very contingency has provided grist to the Hard Problem Mill.
(4) There’s a vast amount of literature on these issues in analytic philosophy. And that includes a large number of competing claims and competing technical terms for different aspects of consciousness. Indeed, for around twenty years (say, between 1990 and 2010) , the issue of consciousness was already the most fashionable of all.
(5) Admittedly, Mo Costandi might not have have been putting his own position here. What’s more, this passage isn’t an actual definition of the word “consciousness” as such. However, it does hint at what must be part of — or relevant to — any theory of consciousness.
(*) More technical discussions of the Hard Problem can by found in my following essays: ‘Is There a Hard Problem of Consciousness… and of Everything Else?’, ‘The Questions ‘Why is water wet?’ and ‘Why does the physical give rise to experience?’ are bogus?’, and ‘David Chalmers’ Unanswerable Question: ‘Why do I have THIS experience?’’.
No comments:
Post a Comment